We say only what we can support
We do not describe capabilities we have not built or results we have not achieved. Where something is early, uncertain, or still being tested, we say so — including when a plainer claim would serve us better.
Responsibility
The standards we hold ourselves to,
whether or not anyone is watching.
Technology work asks people to trust things they cannot inspect for themselves. That trust has to be earned by conduct rather than claimed in a statement. What follows is what we commit to — written plainly, so that it can be checked against what we actually do.
Our Commitments
Six commitments, and what each one costs us.
A commitment that costs nothing is not a commitment. Each of these occasionally requires us to say something we would rather not, decline work we would rather have, or spend longer than a competitor would.
We correct things before we are asked
When we find that something we published or told a client is wrong, we say so and put it right. We would rather raise an error ourselves than have it discovered.
We build security in, not on
Protection is part of how a system is designed rather than something added once it works. We assume our systems will be probed, and we would rather find the weakness first.
We take responsibility for outcomes
Someone is answerable for every piece of work we deliver. If something we built causes a problem, we do not distance ourselves from it — we fix it and we explain what happened.
We use technology with care
Automation and artificial intelligence are held to the same standard as anything else we build: understood before deployed, reviewed by people who are accountable, and used where they genuinely help rather than wherever they can be applied.
We build for the long term
We take on work we expect to still stand behind in ten years. That shapes what we agree to, how carefully we build it, and how long we commit to supporting it.
How We Work
Deliberate before fast.
Most of what goes wrong in technology work goes wrong early, in decisions made before anyone noticed a decision was being made. We try to slow that moment down.
We establish what a system has to be true to before we build it, and we revisit those foundations rarely. Where a foundation turns out to be wrong, we replace it rather than working around it — everything built above inherits its flaws either way.
We design so that no part of a system is irreplaceable, including the parts we are proudest of. A dependency nobody can substitute is a risk regardless of how well it currently performs.
And we write down why, not only what. A choice recorded with its reasoning can be taught to someone new, questioned by someone sceptical, and corrected when circumstances change. One recorded only as an outcome can be repeated or abandoned, and nothing else.
People
The conditions that make the rest possible.
Disagreement is protected
Anyone here can argue against a decision, at any level of seniority, without it counting against them. An organisation that makes honest disagreement expensive stops being told the truth, usually right when it most needs to hear it.
Expertise outranks position
On a technical question, the person who has done the work carries more weight than the person with the more senior title. Seniority decides what we take on; competence decides how it gets done.
Mistakes are examined, not buried
When something fails, we want to understand it properly — which requires that admitting it is safe. Punishing error only teaches people to hide it, and hidden faults are the expensive kind.
If any of this does not match your experience of working with us, we would rather hear about it than not.